An Action-Based Constraint Satisfaction Algorithm for Planning Problems

Abstract

In recent years, the complex and growing constraints in the planning domain have made constraint satisfaction technology a hotspot in the automatic planning field. In the procedure of planning, the action processing plays a vital role. However, diverse but conflicting action relation is difficult to solve in a typical constraint satisfaction method. This paper we code the planning actions as constraints to make the classic planning model a domain-dependent constraint satisfaction problem and let the domain-specific action constraints help guide the search. Based on the model, a dynamic constraint set and constraint process method is proposed to solve the conflicts between the specific action constraints and general constraints in constraint satisfaction problem. The simulation experiments show that the proposed algorithm, with specific constraints, can effectively reduce the planning time and achieves a better performance over other constraint programmed planners.